| {"train/loss": 0.0596, "train/grad_norm": 2.547844409942627, "train/learning_rate": 2.3952095808383236e-08, "train/epoch": 4.996436208125445, "train/global_step": 3505, "_timestamp": 1726510014.761806, "_runtime": 2773.28347492218, "_step": 10, "eval/loss": 0.13957327604293823, "eval/f1": 0.8316917954223653, "eval/precision": 0.8305416576027844, "eval/recall": 0.8328451230913057, "eval/accuracy": 0.9604729231827337, "eval/runtime": 75.264, "eval/samples_per_second": 149.102, "eval/steps_per_second": 18.641, "train_runtime": 2491.7461, "train_samples_per_second": 90.069, "train_steps_per_second": 1.407, "total_flos": 4.319274218928017e+16, "train_loss": 0.1366414232022752} |